I hasten to disagree with you. What else can be guided by, if not the history of the algorithm on historical data?! Otherwise, all traders are like blind kittens.
Before developing any trading system, a system trader finds dependencies and triggers, which in the future give him an advantage in trading on a forward period. Yes, and with your post you question the works of the father of the theory of algorithmic trading, Robert Pardo, with his main work "Development Testing Optimization of Trading Systems". I advise you to read it, very informative!